论文部分内容阅读
为了建立基于遥感影像和环境因子的森林碳密度估测的有效方法,本文基于2009年森林清查数据和SPOT遥感影像,以山西省阳泉地区为例,采用生物量换算因子连续函数法对研究区乔木林地上生物量和碳密度进行估算;在此基础上,选取遥感影像、环境因子(海拔、坡度、坡向等)为自变量,利用增强型BP神经网络建立研究区乔木林碳密度估算模型并输出空间分布图。结果表明:阳泉地区乔木林生物量为552774 t,碳密度为11.38t·hm-2;从不同林型、林龄和起源的生物量及碳密度来看,针叶林、幼龄林、人工林的生物量最大,阔叶林、成熟林、天然林的碳密度最大;采用增强型BP神经网络可以很好地模拟乔木林碳密度,针叶林、阔叶林、针阔混交林仿真结果的平均相对误差和平均相对误差的绝对值分别2.40%、6.87%、-4.09%和6.83%、2.77%、3.99%;基于BP神经网络模型输出乔木林碳密度空间分布图,模拟精度达到85.05%,进一步验证了人工神经网络能为森林碳密度提供快速准确的估测,为今后的森林资源调查和管理提供了科学依据。
In order to establish an effective method for forest carbon density estimation based on remote sensing images and environmental factors, based on the forest inventory data and SPOT remote sensing images in 2009 and the Yangquan area in Shanxi Province as an example, using the biomass conversion factor continuous function method, On the basis of this, remote sensing images and environmental factors (altitude, slope, aspect, etc.) were selected as independent variables, and the BP neural network was used to establish the estimation model for the carbon density of arbor forest Output space distribution. The results showed that: the biomass of arbor forest in Yangquan area was 552774 t and the carbon density was 11.38 t · hm-2. According to the biomass and carbon density of different forest types, ages and origins, the coniferous forest, young forest, artificial The biomass of forest is the largest, and the carbon density of broad-leaved forest, mature forest and natural forest is the largest. The enhanced BP neural network can simulate the carbon density, the simulation results of coniferous forest, broad-leaved forest and coniferous-broadleaved mixed forest The average relative error and the average relative error were 2.40%, 6.87%, -4.09% and 6.83%, 2.77% and 3.99%, respectively. Based on BP neural network model, , Which further verified that artificial neural network can provide a fast and accurate estimation of forest carbon density and provided a scientific basis for future forest resources investigation and management.